I see this as a major issue with the API. Requests with the API are taking me 50 seconds on average unless I significantly cut down on token usage, which defeats the purpose of many applications. The same exact response from ChatGPT took me 10 seconds on the website with a GPT Plus subscription, while it took me nearly 50 seconds on the API. Looking into it, it seems like GPT Plus using text-davinci-002. But this has a 2000 token limit rather than a 4000 one, so how is this possible?
#Chat GPT's API is significantly slower than the website with GPT Plus
36 messages · Page 1 of 1 (latest)
It seems a common problem these days. Chatgpt api indeed works very slow. but it used to be very fast. It might be because the server is over loaded. Just wait for updates
Same is happening to me. From very slow to just not running at all and sending back an error: error: {
message: 'That model is currently overloaded with other requests. You can retry your request, or contact us through our help center at help.openai.com if the error persists. (Please include the request ID 2a1dde40f00c8462afdbfa0ff69cb1af in your message.)',
I don't mind paying for my api usage, but if it is useless... Then what's the point? I think openai need to start think about a downloadable solution for those of us that just want to make dedicated services and not rely on internet at all.
same here
same here
same here
Same here, response time seems to have doubled overnight vs yesterday
getting the same problem today, has anyone found a workaround?
same, trying to use gpt3.5 but it takes minutes sometimes
same here
I thought switching to stream: true would at least give the first few words faster but for me it was the same...
Anyone tried stream: true and had success?
I have, stream working as of this evening. It's trickier to work with, but does send token by token. Huge speed up for me.
I'm using js, but I think there's documentation for a few languages (and some comments on the GitHub that helped)
Which model are you using?
Also, it would be nice to see your minimal workable solution
Sure thing, I'm using 3.5 turbo and I used an implementation here: https://github.com/openai/openai-node/issues/18
The comment all the way at the bottom by ponytojas seems to have an updated version of a previous answer that I was using
Interesting. So this is not using the openai library?
Seems just doing just a post request
I'll try this out
Oh right - that's correct
I was able to incorporate that solution. Thanks for the headsup!
I've been using to explain solution on my beginner friendly coding platform https://pypup.com/problems/add-numbers you can try to select one of the solution to see the results
Hello hello guys, I'm a plus open Ai and want to add the chat gpt in my discord.
would you be so kind to address me to a good link-post- tutorial- etc.
Thank You!
I'm getting timeouts on simple queries that would come back in a second a few days ago. Now it's >20s per request
Same here! I hope that in the future the API will be stable on response times. For production projects that rely heavily on this API, you'll have to trust their server
Hello hello guys, I'm a plus open Ai and want to add the chat gpt in my discord.
would you be so kind to address me to a good link-post- tutorial- etc.
Thank You!
Please don't comment on unrelated threads, but to get started, you can find a github project for a GPT3 discord bot
Concur with obervsations that API calls that were quite fast early last week are now sluggish and that api status shows service degraded or down more often. I assume they are painfully aware of these issues and working as fast as they can to mitigate. Practicing patience.
Thank you for the patience!
I guess I will be following. API calls to 3.5-turbo are taking well over a minute per call. Would other models be faster?